Restricting Noun Incorporation

RESTRICTING NOUN INCORPORATION
Alana Johns, University of Toronto (ajohns@chass.utoronto.ca)

Jan. 2003
ABSTRACT. This paper argues that the phenomenon of noun incorporation in Inuktitut derives
from the fact that the particular set of verbs involved are all light verbs in the sense of verbal
elements excluding the verbal root (Harley 2001). Light verbs in Inuktitut in little v Merge with a
nominal complement as Root. That these elements are light verbs predicts that they are a finite
class within the language with a restricted and distinct semantic range. Parallel data are found in
Salish languages which also have verbal suffixes obligatorily appearing with nominals (Sapir 1911;
Gerdts and Hukari 2002). Rather than adopting a grammaticalized account of the presence of these
verbs (Mithun 1997, 1999), it is argued that these light verbs are members of a universally
available set. Their comparative stability as a class within any one language is explained by this
fact (Butt and Lahiri 2002). Thus noun incorporation in Inuktitut, its limited range of verbs, and its
obligatory nature are all explained readily under a light verb account. This analysis also provides
insight into light verbs as a verb class crosslinguistically, as the delineation of the class in Inuktitut
is quite robust.
INTRODUCTION.
The phenomenon of noun incorporation has received much attention in the linguistic
literature over the years (Baker 1988; 1996, Rosen 1989, Anderson 2000, etc.). A
construction is usually considered to be an instance of noun incorporation when a
(usually bare) noun is found either morphologically attached (1a,b) or in close
association with the verb (1c).
(1) a Wa-ke-nkt-a-hnnu-
Fact-1sS-bed-buy-PUNC1
I bought the/a bed.
b. qukiuti-taar-tunga
rifle-get-intr.part.1s.
I got a rifle.
(Mohawk, Baker 1996 p. 279)
(Inuktitut - Mittimatalik2)
1Abbreviations
for Mohawk are PUNC: punctual. For Niuean they are Abs:
absolutive and Emph: emphatic. For Halkomelem they are aux: auxiliary; det:
determiner. For Urdu they are Perf: perfect. Abbreviations for Inuktitut are intr:
intransitive; trans: transitive part: participial mood; indic: indicative mood; conj:
conjunctive mood; caus: causative mood (usually meaning because or when); mod:
modalis case; neg: negative; rel: relative case.
2Inuktitut examples are either from my own fieldwork, joint work with Jean
Briggs or from published sources. The joint work with Jean Briggs is on the
Utkuhikhalingmiut dialect. The facts are essentially the same across dialects except
where noted. See also Fortescue (1983); Sadock (2002). The orthography used is the
roman orthography of each dialect. The ICI roman is quite similar to IPA. The
Labrador orthography differs in that e, o and are the symbols for the long vowels ii,
uu, and aa respectively. Also capital K is used instead of q.
1

c.
Takafaga
ika
tumau ni a
hunt
fish always Emph Abs
He is always fishing.
ia
he
(Niuean, Massam 2001 p. 157)
In the Mohawk example in (1a) the noun stem nkt bed is embedded within the verb
complex. In the Inuktitut example(1b) the noun qukiuti rifle appears attached to the
left of the verb. Finally in the Niuean example the noun ika fish is found in noncanonical object position (Niuean is a VSO language) adjacent to the verb and
without any case marking.
What has interested linguists over the years is that noun incorporation seems to
involve an argument of the verb, usually the object, which is in close proximity to the
verb and is normally devoid of nominal inflection, e.g. case, number, etc. Thus the
nominal appears on one hand to be a syntactic argument but on the other hand it
appears to be contained within the verbal predicate itself.
Baker (1988; 1996) analyses such constructions as involving syntactic
movement of the head noun from object position so that it adjoins to the verb head. In
contrast, Rosen (1989) argues that a lexical (non-movement) analysis can provide us
with two possibilities. The first of these is where the incorporated noun satisfies
argument structure through word-formation, thus preventing a syntactic object from
appearing. The second is where the incorporated noun acts like a classifier, restricting
the interpretation of the syntactic argument which is external to the verb.
Thus the central issue which has been debated over recent decades is whether or
not noun incorporation is a lexical or syntactic process. To a certain extent the lexical
vs. syntactic issue has disappeared with the emergence of alternative theories such
as Distributed Morphology (Halle and Marantz 1993; Marantz 1997), where
morphology and the syntax can interact.
Neither Baker nor Rosen address Inuktitut in any depth, and both suggest that
Inuktitut noun incorporation may fall outside their general analysis. In fact, Sapir
(1911, 254) states that Inuktitut noun incorporating verbs (or denominative verbs as
he terms them) are not canonical incorporating verbs since the verbal elements are
not verb stems but the verb-forming affixes.... Thus in both Bakers and Rosens
accounts, noun incorporation in Inuktitut is not central to the incorporation debate
and may involve different issues.
Following Sapir (1911), Johns and Massam (1998) argue that noun
incorporation across languages is not a unitary construction (see also Gerdts 1998).
Instead the verb + noun sequences which are termed noun incorporation in the
literature are often syntactically very different, some involving heads and others
involving phrases. What these constructions all have in common, however, is that the
incorporated noun differs in position and/or functional categories from a canonical
object DP.
My central goal in this paper is to demonstrate that noun incorporation in
Inuktitut is the result of the inherent properties of the specific verbs which are
involved. It is these properties which make noun incorporation in Inuktitut distinct
from the cases discussed in Baker (1988; 1996); Mithun (1984); Mithun and Corbett
(1999), etc. More specifically the leading insight will be Sapirs comment mentioned
2

above that the verb-forming affixes are not verb stems. That these elements are
not standard verbs stems will be shown to explain a wide variety of properties
associated with the construction, including the fact that they are a finite set and have
very general meanings (Johns 1987; Mithun 1999).
The nature of these verbal elements leads us to another set of familiar issues in
linguistics, but one which is not usually found in a discussion of noun incorporation.
This is the existence in languages of so-called LIGHT VERBS or verbs which must
combine with another element, often a nominalized form of a verb, to form a complete
predicate (Jespersen 1965; Kearns 1988; Grimshaw and Mester 1988; Butt and
Lahiri 2002). Although the complement of a light verb is usually considered to be
either verbal noun, as in English take a rest, I assume here a more general definition of
light verb, such that light verbs may also take nominal (non-verbal) complements
(Harley 2001). What makes these verbs light in both instances is their lack of
semantic complexity plus their syntactic dependency on another element as ROOT
(Pesetsky 1995; Marantz 1997).
Analysis will show that the basis for noun incorporation in Inuktitut is directly
linked to the class of the verbs involved.3 More specifically, these verbs are all light
in the sense that they do not contain any semantic elements which characterize full
verbal entries, especially manner (see Ritter and Rosen 1997 for an analysis of the
verb have in English along these lines). Instead light verbs consist solely of a varying
number of verbal operators (see Koenig and Davis 2001 for operators within lexical
items), e.g. negation, etc. It is the lack of root properties of these verbs which requires
that they merge with noun roots. The overall thrust of the analysis here is to identify
and categorize the empirical restrictions of the phenomenon in Inuktitut we call noun
incorporation, and to show at the same time that the properties underlying these
restrictions are the basis of the entire phenomenon itself.
1. NOUN INCORPORATION IN INUKTITUT
Noun incorporation (or NI) in Inuktitut is very productive and very common in the
language.4 While the noun is morphologically attached to the verb, it nonetheless has
properties which indicate that it is not merely a verbal modifier nor a compound5. In a
seminal article Sadock (1980) shows that in Kalaallisut (West Greenlandic Inuit), NI
3For
one important view on verb class, see Rappaport Hovav and Levin
(1998).
4Parkinson (1999) shows that Inuit children acquire noun incorporation at an
early stage. This fact is expected under the account given here, given the simplicity
and universality of the verbs.
5There are a number of pieces of evidence which make a compound analysis
impossible. First and foremost, Inuktitut does not have either verbal or nominal
compounds, i.e. there is nothing equivalent to English toothbrush or photocopy. See
Sadock (1980; 1991) for arguments against compounding.

is clearly syntactic (to be discussed in 1.1), rather than simply compounding.6 Thus
noun incorporation in Inuktitut provides an interesting example of the interaction
between the syntax and the morphological component.
1.1. Previous Investigations of Noun Incorporation in Inuktitut
Noun Incorporation in Inuktitut has been discussed by a number of linguists over the
years. Undoubtedly the most influential paper is that of Sadock (1980) who labels the
Inuktitut construction noun incorporation and demonstrates many of its syntactic
properties. One the most striking of these is the potential for the incorporated noun to
have referential properties. Sadock shows that nouns in noun incorporation
structures can be referred back to in the discourse context. Using examples such as
(2a), from a Kalaallisut childrens story book, he shows that once an incorporated
noun has been introduced into the discourse, subsequent sentences may refer to it
through agreement on verbs. Similar examples are found in the Canadian dialect
Mittimatalik, as in (2b). I have coindexed the relevant elements to aid in
interpretation.7
(2)a. Suulut timmisartui-lior-poq
plane-make-intr.indic.3s.
Sren
Sren made a planei
Suluusa-qar-poqi
aquute-qar-llu-nii-lu
wing-have-intr.indic.3s.
rudder-have-inf.-3Rs-and
Iti has wings and a rudder (Kalaallisut from Sadock, 1980; 311)
b. Johnny uvirnirui-liu-laur-mat
Johnny shirt-make-past-intr.caus.3s.
Johnny made a shirti
nulia-nga
angirra-rami
taku-llu-ni-uki
wife-Poss3s.
home-cause4s.
see-conj.-4s.-3s
And his wife came home and she saw iti
(Mittimatalik)
6That
noun incorporation in Inuktitut is not tied in with general word formation

processes makes it distinct from noun incorporation in Iroquoian languages. Mithun
and Corbett (1999) discuss the fact that the incorporated noun and its verb in
Mohawk often mean more than the sum of the parts, in effect having a lexicalized
status. To my knowledge, this never happens in Inuktitut, with the possible exception
of body part ailments - see section 2.4.2.
7Note that Inuktitut does not distinguish 3 person for gender or animacy.
Accordingly the translations are given in English according to context, or translation
of the moment.
4

Thus we see in (2a) that the third person singular agreement markers -poq and -ni in
the second line refers back to the incorporated noun timmisartu- plane in the
proceeding sentence. Likewise in (2b) the third person singular agreement elementuk in the second line refers back to uvirniruq shirt.
Crucially however, although incorporated nouns may be referential, they are not
always referential. This fact is not stated explicitly either by Sadock (1980) or Van
Geenhoven (1998a, b), who investigates extensively the referential properties of
incorporated nouns.8
(3) *ilinniatitsi-u-junga
quviasuk-tuq
teacher-be-intr.part.1s. happy-intr.part.3s.
Ii am a teacher. He/shej, *i is happy.
(Iqaluit)
As in English, the nominal introduced by the copula cannot have reference

independent of the subject. Also like English, the existential properties of the noun
depends on the properties of the individual verb, e.g. have a bird vs. be a bird. This
distinction sets the stage for the general thrust of this paper, which is that inherent
properties of verb are the determining factor of noun incorporation.
Another important observation in Sadock (1980); (2002) is that a subset of
incorporating verbs allow possessive inflection on the noun root, while the rest do not.
This can be seen in (4).
(4) a. Kalaall-it
nuna-a-liar-poq
Greenlander-pl. land-3pl.-go-intr.indic.
He went to Greenland (i.e. to the Greenlanderss country)
(Kalaallisut: Sadock 1980; 314)
b. sugusi-up
illu-nga-no-vunga
child-rel.
house-3s.-go.to-intr.indic.1s.
I am walking to the childs house
(Labrador Inuttut)
I will not be addressing this class of NI verbs, pointing out only that this distinction is
clearly along the lines of verb class, and that the crucial factor which is involved is
that the verb is locational/directional (Sadock 2002).9
8Van Geenhoven locates the source of the referential properties of the noun in
the verb rather than the noun itself but does not discuss difference across verbs in
this regard. In fact, her statement in Van Geenhoven (2002, 766) that a referentintroducer or, as I called it, as a semantically incorporating verb... might lead the
reader to assume that they are equivalent.
9Interestingly, these verbs are the only ones which allows genuine possessor
stranding in Canadian Inuktitut. Canadian dialects do not allow possessor stranding
unless there is possessive inflection on the head noun. Sadock (1980; 1991) shows
that possessor stranding without possessive inflection is possible for certain

Sadock (1985; 1991) argues that noun incorporating verbs are both
morphological and syntactic at the same time. Within his Autolexical Theory of
syntax, this is not a problem since Autolexical Theory allows both morphological and
syntactic representations, which have independent but connected representations, so
long as the representations do not conflict in their mapping.
Bok-Bennema and Groos (1988) follow Sadock by adopting an analysis where
the noun incorporating verbs in Kalaallisut are both morphological and syntactic at
the same time. They propose a very simple approach to noun incorporation in the
Inuit language, which is that the noun is linked to the verb by a left-adjacency
requirement of the verb. This approach prefigures the syntactic account adopted here
and later proposed in Van Geenhoven (1998a, b; 2002), Massam (2001) which is that
the noun is base-generated in its position. Specifically I follow Hale (2001) who claims
that it appears there through the operation Merge (Chomsky 2001). Under BokBennema and Grooss account, there is an operation similar to Merge but the
features which trigger it are idiosyncratic, i.e. any verb which has affixal properties
will participate. Crucially the proposal argued in this paper is that the class of verbs
which has these properties is predictable on the basis of the lexical semantics and
category of the verbal element. Thus a natural class of verbs underlies and
determines the nature of the noun incorporation in Inuktitut.
Recently Van Geenhoven (1998a, b) presents an extensive investigation into
the formal semantics of nouns involved in incorporation in Kalaallisut. She proposes
an account involving semantic incorporation, whereby semantic properties relating to
indefinites are absorbed by the incorporating verb and the incorporating verb itself
conveys the existential interpretation of the nominal. Thus, similar to all previous
analyses, incorporating verbs are considered to be identical to non-incorporating
verbs except that they have some additional property. For Bok-Bennema and Groos
it was a morphological requirement; for Van Geenhoven, the verbs have absorbed
something extra, i.e. they are by definition semantically more complex. This is the
exact opposite of the analysis proposed in this paper. In the sections to follow, I will
show that the semantics of the Inuktitut incorporating verbs is considerably less
complex than that of non-incorporating verbs, and it is this simplicity, or lightness,
which is the actual property underlying the phenomenon of incorporation in this
language. Thus no additional properties must be posited to account for incorporating
verbs as a class. In fact it is the paucity of verbal properties which creates the need
for morphological attachment (to a nominal).
While previous analyses have shed much light on noun incorporation in
Inuktitut, there is one prominent omission in the discussion overall. This is the fact
that, although every linguist is aware that NI in Inuktitut is based on a certain
subset of verbs, no one has provided any explanation for the questions which arise
from this fact. These are: a) Why is noun incorporation obligatory when it is possible?
b) Why is it restricted to a certain set of verbs? and c) Why is it restricted to these
particular verbs and not others? As noted in Johns (1987), the lexical semantics of
lexicalized possessor/possessum combinations in Kalaallisut. It is a common
misconception that this is a widespread phenomenon.
6

incorporating verbs appears to have a commonality. After a brief exposure to the
class, a linguist can make fairly good predictions as to which verb meanings cannot
be incorporating verbs, although it is more difficult to predict which ones will be found.
In the analysis to follow, I propose a basis for our understanding of this verb class as
a whole. I will show that incorporating verbs in Inuktitut all consist of incomplete or
light verbs, i.e. verbs lacking root elements (Marantz 1997).
2. THE INCORPORATING VERBS IN INUKTITUT
Noun incorporation in Inuktitut is restricted to a finite set of verbal items within each
dialect. The set of verbal items is generally consistent across dialects, although the
members of the set can vary slightly. Unlike other languages where NI appears to be
optional, e.g. Mohawk,10 if a verbal element may appear without an incorporated
noun, as in (5a) illustrating the verb nigi- to eat, then an N can never be
incorporated, as illustrated by (5b).
(5) a. pitsi-mik
nigi-vunga
dried.fish-mod.
eat-intr.indic.1s.
I am eating dried fish
b. *pitsi-nigi-vunga
dried.fish-eat-intr.indic.1s.
(Labrador Inuttut)
(Labrador Inuttut)
Equally, if a verbal item permits NI, as in (6a), then there must obligatorily be NI, as
shown by the ungrammaticality of (6b), where the NP is externalized in a syntactic
fashion parallel to the grammatical (5a) above.
(6) a. pitsi-tu-vunga
dried.fish-consume-intrans.indic.1s.
Im eating dried fish
(Labrador Inuttut)
b. *pitsi-mik
dried.fish-mod.
tu-vunga
consume-intr.indic.1s.
(Labrador Inuttut)
Even though pitsik dried fish has modalis case attached in (6a), as in (5a), the lexical
item -tu- consume (often translated as eat) cannot appear alone. When no nominal
is available, a dummy element pi- must attach to the verb, as in (7)
10Baker (1988; 1996) treats noun incorporation in Mohawk as optional but

Mithun and Corbett (1999) argue that it is a lexicalization process and therefore the
incorporated form and the non-incorporated form are not equivalent.

(7)
pi-qa-nngit-tuq
null-have-neg.-intr.part.3s
He has nothing [Literally: He does not have anything.]
(Iqaluit)
Thus it is clear that in Inuktitut there is a major division between verb classes, one
class must incorporate nominals and the rest (the elsewhere case) cannot. It is also a
well-known fact that the incorporating class is a smaller closed set of verbs
(numbering around one hundred based on Fortescue 1983) than those which do not
incorporate (number unknown). This difference is reflected in the fact that while the
larger class of non-incorporating verbs are found in general dictionaries, the smaller
set of incorporating verbs are found in what are called postbase dictionaries (noninflectional affix dictionaries) or subsections of dictionaries (see for example Lowe
2001).
One should also note that there are no restrictions as to which nominal may be
incorporated other than general pragmatic ones. Thus how commonly a noun is found
associated with the verb has no bearing on the likelihood of it being incorporated,
unlike the cases described for Iroquoian by Mithun (1999). The source of the
incorporation stems only from the verbs requirements.
Syntactic incorporation analyses such as Baker (1988; 1996), which are based
on the idea of movement do not lend themselves to the sort of construction we
observe with Inuktitut NI. As movement entails either optionality or alternation with
a non-moved element, movement runs counter to the facts here. In short, a different
type of analysis is called for, especially one which will explain the fact that the
incorporating verbs are a finite set while the non-incorporating verbs are an open set.
As mentioned above, all previous investigations of Inuktitut noun incorporation to
date have ignored the issue of which verbs incorporate, focussing instead on either the
syntactic or formal semantic properties of the construction in and of itself, and
especially properties of the incorporated nominal.
2.1 A Lexical-semantic Analysis of Noun-incorporating Verbs in Inuktitut
Johns (1999a, 2000) pointed out that there are two related properties not explained
by syntactic, lexical/compounding or semantic accounts of NI in Inuktitut. These are
a) that the NI construction seem to involve very basic verbs with similar semantics
and b) that verbal elements in NI constructions have a more general meaning than
those in non-NI constructions. The explanation proposed is that NI in Inuktitut is
constructed with a type of light verb which must take a nominal root to form a
complete structure.
Mithun (1999, 50) also points out that in Yupik, a language quite closely
related to Inuktitut, not only do verbs and verbal affixes differ in distribution but that
the affixal variety (i.e. incorporating verbs) are typically more general and diffuse
semantically than roots. It is exactly this type of generalization which is being
utilized here in order to explain incorporation (although from a different perspective

than that in Mithun 1999).11 I will be arguing like Mithun that the semantics of
incorporating verbs is less complex than that of non-incorporating verbs. This
contrasts with all previous analyses mentioned above.
In Van Geenhoven (1998a, b), existential operators are contained within
incorporating verbs. Her central concern is to argue that the incorporated N is a
predicate, and that incorporation involves base-generation and not movement of the
nominal. Van Geenhoven distinguishes between incorporating verbs and nonincorporating verbs, as shown in (8) and (9), where I provide an Inuktitut example in
a. and the corresponding formal semantic representation of that particular verb from
Van Geenhoven (1998b) below it.
(8) a. Non-incorporating eat
puiji-vini-mmik
nigi-vunga
seal-former-mod.
eat-intr.indic.1s
Im eating some seal meat
b. ye xe [eat(x,y)]
( 9). a. Incorporating eat
puiji-vini-tu-vunga
seal-former-consume-intr.indic.1s
Im eating seal meat
b . P <e,t> xe y[eat (x,y) P(y)]
(Labrador Inuttut)
(Van Geenhoven 1998b, ex. 38)
(Labrador Inuttut)
(Van Geenhoven 1998b, ex. 32)
We see then that Van Geenhoven does distinguish between the two verb types in (8)
and (9). However, the formal semantics of the incorporating verb eat in (9) is more
complex than that of the non-incorporating eat in (8). In fact, the semantics of the
incorporating verb subsumes that of the non-incorporating verb. The incorporating
verb has a number of extra properties, including a) a slot for the property denoted by
the nominal plus b) the incorporating verb contains an existential as part of the verb
meaning. In terms of complexity of verb types I will propose something very different,
in fact the opposite. I argue that the non-incorporating verb eat is less specified with
verbal meaning, and hence is not a full verb but is instead a light verb.
As mentioned above a problem for all previous analyses including Van
Geenhovens is that they do not directly address the issue of which verbs incorporate.
The question remains open as to why there are no NI verbs like break, tickle, cook
etc. with the properties shown in (9) above. There is nothing to prevent such cases.
Yet such verbs never incorporate. The closest Van Geenhoven (1998b; 243) gets to
11Mithun claims that NI is used in instances where the action is typical, and is
less likely to be used in novel situations involving the N. I have seen no evidence of
this distinction in Inuktitut.

this question is her statement a semantically incorporating verb does not introduce
stages of a kind.12 This suggests that verbs where the object undergoes a change of
state never incorporate. Instead we observe that incorporating verbs can describe an
object coming into being, coming into sight, etc. but they cannot describe an object
changing from one state to another. We also observe there is never any description of
the manner by which the subject of the incorporating verbs comes into contact with
the object, thus tickling, brushing, etc. are not possible incorporating verbs. None of
these possibilities is ruled out by the analysis in (9).
What we find upon examining NI verbs as a class is that they generally seem
to involve either a predicate nominal or something to do with an objects existence. All
other NI verbs are simply variations of these two basic types, the variations brought
about by the presence or absence of operators of negation, etc. This simplicity of
class explains why many incorporating verbs are antonym pairs, e.g. getting/losing,
etc. What is needed is a theory of verb class in Inuktitut where the incorporating
verbs do not contain canonical verbal material, but instead the most minimal verbal
elements along with predictable variations. In the following sections I will provide an
outline of this verb class in Inuktitut, showing how NI verbs display the properties of
light verbs (v), i.e. verbs without root () elements (see Pesetsky 1995; Marantz
1997). Thus the main basis for the claim that Inuktitut incorporating verbs are light
verbs is based on the argument that all of the member of this class may be readily
characterized as light verbs. There can be no exceptions. Support for this claim that
noun incorporation in Inuktitut is a light verb plus nominal root comes from Harley
(2001) who claims light verbs without verbal roots take nominal complements. In the
next section, I will survey the class of noun incorporating verbs, and show that every
single one is a light verb.
2.2 NI Verbs in Inuktitut
The Inuktitut word obligatorily consists of a root and an inflection. Between these two
elements appear a potentially infinite number of affixes, often referred to in the
literature as postbases (or non-inflectional affixes). Postbases consist of verb
modifying elements, nominal modifying elements, elements which change a verb into
a nominal and elements which change a noun into a verb. The latter are the noun
incorporating affixes. The most extensive and important research into the
distribution and properties of postbases is Fortescue (1983). Fortescue divides
postbases into 26 classes on the basis of semantic similarity. Among the many
postbases are those which appear to be verbs by virtue of their English translation.
However Johns (1999b) argues that volitional verbs are actually modals and, as
such, are not full verbs.
Of Fortescues 26 classes of postbases, nine contain noun incorporating verbs.
12 In fact Van Geenhoven (2002, p. 805) explicitly states that incorporated
nouns are predicative restrictions on the implicit argument of the verb. This goes
against the very essence of the present analysis which is that incorporating verbs do
not take arguments at all.
10

Note that Fortescue himself does not describe this phenomena as noun incorporation
but as verb forming.13 Fortescue provides approximately 90 examples of different
incorporating verbs from each of four dialects, ranging across the arctic from
Greenland to Alaska.14 The affixes are remarkably similar in both meaning and form
across dialects (once the phonological systems of the different dialects is taken into
account).
In (10) I provide Fortescues nine classes, each followed. Note that the class
labels refers to a semantic commonalities; thus in Fortescue (1983) each class
represents a set of related NI verbs.
(10)a. Being and becoming
arna-u-junga
woman-be-intr.part.1s
Im a woman
(Mittimatalik)
b. Lacking
ulu-iruti-junga
ulu-lack-intr.part.1s.
Im out of ulus [ulu is a womens knife]
(Mittimatalik)
c. Feeling
siu-siri-juq
ear-trouble-intr.part.3s
His ear aches/hes having trouble with his ear
(Mittimatalik)
d. Having
savi-qaur-tunga
knife-have.a.lot-intr.part.1s
I have plenty of knives
(Mittimatalik)
e. Acquiring
qukiuti-taar-tunga
I got a rifle
(Mittimatalik)
13The
reader might posit that if one were simply to claim that these elements
were affixes, the analysis would fall out. I disagree with such an analysis because it
would lose the generalization shown here that it is the nature of these verbs which
makes them affixes, not vice-versa. Counterevidence to my claim would be the
existence of an independent copula in Inuktitut. The affixal analysis is unfalsifiable.
14These are West Greenlandic, Tarramiut, Copper and North Slope.
11

f. Movement
sugusi-up
illu-nga-no-vunga
child-rel.
house-3s.-go.to-intr.indic.1s.
I am walking to the childs house
(= ex. 4b. above)

(Labrador Inuttut)
g. Acting and seeming like

Naatali-urquuji-juq
Natalie-resemble-intr.part.3s
He/she looks like Natalie
(Mittimatalik)
h. Doing with and providing

tuktu-vinir-tur-tunga
caribou-former-consume-intr.part.1s.
Im eating caribou meat
(Mittimatalik)
i. Judging and Saying15

tipatsauti-sunniq-tuq
perfume-smell-intr.part.3s
It smells like perfume
(Iqaluit)
I propose that the basis of each NI verb is a light verb, and that variations among
the set them involve either simple modification through addition of modal elements
from a fixed set, e.g. negation, or other elements of this type. In other words, the verb
classes consist only of what in Generative Grammar is called little v. No V or root
segment is present. The proposal argued for here resembles in some respects the
analysis in Koenig and Davis (2001) who, working in a HPSG framework, claim that
the lexical entry of a verb is divided into two parts. One part is the modal part (here
little v), which can be modified by semantic operators, such as negation, etc. The
second part is the situational core (here the root), which contains the verbs core
argument structure. What is novel in the claim presented in this paper is the claim
that there is no V element (or situational core) involved in incorporating verbs;
instead these verbs obtain a root through Merge with a nominal (see Harley 2001 for
an example of such an analysis of the English verb write.). In principle then, the exact
subset of verbs which incorporate in Inuktitut is not just a problem to be solved, but
is instead a robust clue to the nature of verbal category.
Following Koenig and Davis (2001), I will be assuming that there exists some
fixed set of operators which may be found in the modal component of a lexical entry.
In theory any one of these operators may combine with any one of the others, but
there is clearly a limit on the number which may be present within any one verb. This
restriction immediately limits the class of incorporating affixes, thus leading to a
finite set, which is the case. The modal elements involved are familiar semantic
15This class is mixed in that it contains both verbs which are nounincorporating as well as verbs which attach to clauses or quotations, e.g. niraq say
(that).... I will only deal with verbs from this group which attach to nominals.
12

primitives, e.g. negation, quantity, etc. As a result of this restriction to basic
elements, the majority of incorporating affixes are copulas, verbs of appearance, etc.
The claim here is therefore that there is no genuine verb per se in incorporation
constructions in Inuktitut, and that all the incorporating verbs are based on the
equivalent of a semantically minimal verbal core (the light verb), with permutations
created by the addition of the operator elements. In the following I will present an
overview of the major elements which contribute to the meaning of incorporating
verbs in Inuktitut. As expected within such a proposal, we will see is that these verbs
frequently come in sets, related through relations of antonymy, the physical senses,
and quantification, etc.
2.3 Operators and Light Verbs
In this section I will outline some of the basic semantic operators which are found in
incorporating verbs. The main contrast between incorporating verbs is whether or
not the incorporated noun has independent reference or not. This is the basic
difference between be and have. Following Van Geenhoven (1998a, b), I will be
arguing that it is the verb itself which establishes the contrast, not the nominal. I
differ from Van Geenhoven, however, in that I will be assuming that Inuit NPs are
inherently existential, thus simultaneously explaining the lack of determiners. As a
result, in the possessive and existential examples in (11), each nominal comes with its
own existential operator. The light verb in question is semantically empty, as shown
in (11c).
(11) a. qimmi-qaq-tunga
dog-have/exist-intr.part.1s.
I have a dog.
(Iqaluit)
b. tuktu-qaq-tuq Nunavu-mi
caribou-have/exist-intr.part.3s.
There are caribou in Nunavut.
(Iqaluit)
c. [ ]
Van Geenhoven (1998a, b; 2002) does not explain why the process of incorporation
does not produce an existential interpretation for copulas and related incorporating
verbs, as seemingly predicted by her analysis. The question for the present analysis
is why is it that the existential interpretation of the nominals in these constructions
seems to disappear, as in (12).
(12) a. Saali ilisaiji-u-juq
Sally teacher-be-intr.part.3s.
Sally is a teacher.
(Mittimatalik)
13

b. inu-u-runga
inuk-be-intr.part.1s.
I am an Inuk (Eskimo)
(Utkuhikhalingmiut)
c. [ I ]
If ilisaiji teacher comes with an existential operator by default, there must be
some mechanism whereby it can be explained that there are only two entities in the
sentence. Here I appeal to the notion of predication or coindexing through operators.
The copula contains something which the possessive does not, an identity operator
linking the two existentials - that of the incorporated nominal and that of the subject.
Thus the lexical semantics of the copula is shown in (12b).
In summary, the possessive and the copula are minimally distinguished by the
introduction of an identity operator. The similarities between copulas and verbs of
possession has long been noted (see discussion in Ritter and Rosen 1997).
2.3.1 Negation
We have seen the first and major division between incorporating verb types involves
whether or not the reference of the nominal is identical to that of the subject of the
verb. It is now easy to see how other verbs form sets based on the simple addition of
an operator. For example, negation may be added to many of the verb meanings,
producing antonyms. Thus, just as something may be possessed (see 11a above), the
opposite is also be found.
(13)a. ulu-iruti-junga
ulu-lack-intr.part.1s.
Im out of ulus.
(Mittimatalik)
b. [ ]
We note that there is no NI verb which is the negation of the copula. That the subject
and the nominal are linked through an identity operator of the copula may explain
this absence. The result would a negation of the existence of both. We will see more
examples of negation, as we examine more incorporating verbs.
2.3.2 Quantity.
Quantity is also an operator which serves as the basis of many of the verbal
distinctions. We will see that it can quantify over the nominal or over the verb itself,
resulting in aspect. Quantify or Q is something like an adverbial clitic. Q can be found
cliticized to the right of the light verb. Since possession is the most basic verb
meaning here, this will result in the meaning to have lots of, as shown in (14).
14

(14)a. savi-qaur-tunga
knife-have.a.lot-intr.part.1s
I have plenty of knives.
(Mittimatalik)
b. [ ]Q
Naturally it may also appear with negation, as in (15).
(15)a. savi-kiksa-rama
knife-not.have.enough-intr.caus.1s
I am short of knives.
(Mittimatalik)
Q
b. [ ]
Note that there is no independent negative element in the verb -kiksaq- not have
enough. The meanings are part of the light verb in question.
2.3.2.1 Quantity and Aspect
The operator Q will also play a role in the aspectual nature of the light verb. Aspect
can be viewed as the quantification of an event, or its measuring out as Tenny
(1987) puts it. Given that light verbs are inherently states, the only effect which Q
can have on these states is to turn them into an event by prolonging them.
Koenig and Davis (2001) posit inchoative aspect as a possible operator within the
modality component of the lexical item (or little v, as we have it here). The inchoative
aspect creates an incremental period resulting in the final state. The verb
-qaq- have/exist is only a state and does not have inchoative aspect; however the
verb -taaq- get does, as shown in (16).
(16)a. qukiuti-taar-tunga
I got a rifle.
(Mittimatalik)
b. Q[ ]
Here we see that the addition of the quantifier to the left of the light verb produces an
inchoative meaning, whereby the having has been turned into an event. Note that
this differs from causative inchoatives where there is some independent entity
causing an event to take place, as in English transitive break.
We are not surprised that a negative form of (16b) exists as an incorporating
verb as well.
15

(17)a. ulu-irsir-tuq
womans.knife-cannot.find-intr.part.3s.
She cant find her ulu.16
(Mittimatalik)
b. Q[ ]
In a similar fashion, the copula can have Q added to it, producing the inchoative (18)
(compare the copula in 12a).
(18)a. kigusiriji-nngur-tuq
dentist-become-intr.part.3s
He/she has become a dentist.
(Mittimatalik)
b. [QI ]
Finally the quantifier can be the sole occupant of the light verb itself, producing
something of an iterative meaning to be occupied or busy with something, as in (19).
(19) a. qukiuti-liri-juq
rifle-do.with-intr.part3s.
He/she is playing with/fixing the rifle.
(Mittimatalik)
b. [ Q ]
c. kiguti-liri-ji
tooth-do.with-agent.nominal
dentist
(Mittimatalik)
As example (19c) indicates, this morpheme is sometimes found in terms denoting

professions.
As can be seen from the above examples, many of the verbs from Fortescue
(1983) fit in exactly with the analysis proposed here, which is that these verbs
consist only of basic semantic primitives, such as negation, etc. We have already
dealt with classes a. (12, 18), b. (13, 15, 17), d. (11, 14), e. (16) and one from h. (19)
from the set of classes in (10) above. Note that Fortescues classes do not map
exactly to the present analysis. While Fortescue has having and acquiring in
different classes (d. and e. respectively), he also puts lack and losing in the same
class b. Here, class derives from the presence of similar operators. There are no de
facto classes. Thus the analogies between having, lacking, getting, and not finding
are more readily captured.17
16In some dialects, the cognate of this morpheme is translated as to lose (see
Fortescue 1983).
17The NI verb -siuq- meaning to look for is clearly related to the meaning
-taaq- get; however it is not obvious whether the best means to formally express this
16

In the next section, I turn to more complicated cases for this proposal (classes
10c., g. and the rest of h.)
2.4 Elaborations on the Basic Cases
In this section we will see that the remaining case of verbs can all be explained on the
basis of the core cases argued for in section 2.3. It is clear that each and every one of
the NI verbs is distinct from canonical verbs which contain verb roots.
2.4.1 Physical Similarities
One group of incorporating verbs (Fortescues class i. from 10. above) all involve the
evaluation of the identity of something through its physical properties. Thus
something is said to seem like X in some physical way. Consider the example in (20).
(20) a. Naatali-urquuji-juq
Natalie-resemble-intr.part.3s
He/she looks like Natalie.
b. qamiuti-qpaluk-tuq
sled-resemble-intr.part.3s
It looks like a sled
(Mittimatalik)
(Utkuhikhalingmiut)
c. [ IVISION ]
Recall that the core distinction between many NI verbs involves whether or not the
nominal has independent existence. The NI verb in (20) straddles this distinction in
that identity is involved but only in a restricted sense. Identity is predicated only
through visual properties, no other. We can use our operator I which is found in the
copula but restrict it to identity only in visual detail, i.e. not complete identity. This is
shown in (20b) where the term VISION will be used as a physical restrictive.18 We
find that a number of verbs pattern in this fashion, as seen in (21-23).
(21) a. urqusaut
sikituur-valuk-tuq
furnace(abs.)
skidoo-sound.like-intr.part.3s.
The furnace sounds like a skidoo.
(Mittimatalik)
b. [ ISOUND ]
relation is to introduce a new operator having to do with intension, i.e. that acquiring
is a mental construct, or to embed the formalism for -taaq- within an activity (Q). In
addition, there remains research to be done on the scope of the meaning of this verb.
18The formalisms here are similar to Jackendoff (1990) and earlier work.
17

(22) a. unnir-sunnir-tuq
armpit-smell-intr.part.3s.
It smells like an armpit.
b. anaq-hungnit-tuq
feces-smell-intr.part.3s
It smells like excrement
(Mittimatalik)
(Utkuhikhalingmiut)
c. [ ISMELL ]
(23)a. Naatali-rjuujaaq-tuq
Natalie-act.like-intr.part.3s.
She is acting like Natalie.
(Mittimatalik)
b. [ IACTION ]
Interestingly, rather than the full five physical senses, we find both touch and taste
missing, yet action is there. In Inuit culture, taste and smell are very closely linked,
so the lack of an independent taste NI verb comes as no surprise. The explanation
behind the lack of touch verbs is probably related to the fact that the verbs of
restricted identity entail that the speaker be some distance from the entity being
evaluated.19 Finally the behaviourial NI verb is likely to be a further restriction of the
visual restriction.
2.4.2 Ailment Verbs
The majority of Fortescues class c. verbs from (10) above fall under the definition of
feeling an unpleasant sensation in some body part over a prolonged period of time.
These verbs are related to the verb -liri- meaning to be occupied or busy with
something in (19) above. The meaning of pain, etc. will derive from the context of the
nominal, which is a body part. We can assume that body parts are in some sense
inalienable universally, and therefore that their existence is dependent on another
existence. With respect to the NI verbs of feeling, the interpretation is that a) that
the nominal is a subset of the subject of the verb, and b) that this subset nominal
occupies the attention of the owner over a period of time. This results in a negative
connotation.
(24) a. siu-siri-juq
[same as 10c]
ear-trouble-intr.part.3s
His ear aches/hes having trouble with his ear
(Mittimatalik)
b. [ Q ]
19This
explanation could also explain why taste is absent (David Johns,

personal communication).
18
These verbs have the same properties as the one in (19b). The difference in meaning
is a result of the subset relation between the nominal and the subject. We note that in
the Mittimatalik dialect, lexicalization has taken place such that the incorporating
verb is not identical to -liri- in (19), nor is the nominal a free lexical item (siuti is the
word for ear). This is the only class of NI verbs where lexicalization seems to have
involved the noun. However Fortescue (1983) gives -liri- (see 19 above) with an
ailment meaning in a number of dialects. He in fact classifies -liri- both as class c.
and h. Here -liri- is considered to be essentially the same morpheme, with meaning
differences deriving from context (see Johns 1999b).
2.4.3 Verbs of Creation and Consumption
The next set of verbs appear to be quite challenging for the approach taken here, as
they involve verbs of creation make and consumption eat. However we will see that
there is once again evidence that these verbs are underspecified for meaning, and
that moreover these very meanings have been claimed to those of light verbs, e.g
Harley (2001). In fact this class of verbs has received its own term within the field of
lexical semantics - incremental theme verbs. Thus we again are not surprised that
creation and consumption are unmarked antonyms of one another, i.e are linked via
negation as we have seen above. However in this case, negation will produce not a
non-event but an opposite process.
The first thing to note about incorporating verbs of creation and consumption in
Inuktitut is that, like all the NI verbs we have seen so far, they specify very little
detail of the action, as shown in (25).
(25) a. tuttu-vini-tu-vunga
caribou-former-consume-intr.indic.1s.
Im eating caribou meat.
b.
tii-tuq-tunga
tea-consume-intr.part.1s
Im drinking tea
(Labrador)
(Iqaluit)
As noted in Mithun (1999, pp. 49-56), Eskimo-Aleut incorporating verbs are generally
broad in meaning. Thus the verb -tur- means consume rather than eat or drink,
since the latter translations derives from the context, as shown by the contrast
between (25a and b). In fact, -tur- is very broad in meaning, with no detail about
mouth movement or any physical action involved. In some ways, this is similar to the
English have a cup of coffee (see Ritter and Rosen 1997), except that the Inuktitut
verb entails ingesting the item in question, while the English have can go with a good
nap, etc. (see also Mithun 1999).20
20In
Utkuhikhalingmiut -tuq- can not only mean eat, drink but also wear,
use (Jean Briggs, personal communication).
19

The creation verb is shown in (26).
(26)
jappa-liur-tunga
parka-make-intr.part.1s
I am making a parka
(Iqaluit)
Again, the action is very underspecified as to the means by which it took place.
Mithun (1999, 50) shows that in Yupik, a language closely related to Inuktitut, -liurcan mean brush (teeth); cook (reindeer meat); chop (wood); work (at store) or cut
(fish), all depending on the context.
Both verbs, as incremental theme verbs, involve the incremental bringing into
existence of a nominal and its converse, the incremental elimination from existence of
an entity. The entries for these verbs is therefore as in (27).
(27)a. -liuq- make
[Q [ QI] ]
b. -tuq- consume
[Q [ Q I] ]
What these representations indicate is that in the case of -liuq- make, there is an
activity, whereby the identity of the nominal becomes gradually non-identical from
that of the subject of the clause, i.e. first there is only one entity, but over time, there
will be two distinct entities. Note that the outside Q resembles the formalism in (19b)
for being occupied with something. In a similar fashion, the representation in (27b)
indicates that the verb -tuq- consume involves an entity busy with another entity
such that the latter becomes identical in identity with the former.21
In summary, we have seen that verbs of creation and consumption fit our
definition of light verbs. They are extremely underspecified in their semantics; they
can be accounted for with simple semantic operators. Significantly, they are
antonyms. That creation and destruction verbs are currently considered to be light
verbs by independent analyses (e.g. Harley 2001) is further support for this proposal.
2.4.4 Polyadic NI Verbs
The remaining type of verbs which fall under the account proposed here are the class
of verbs discussed in Van Geenhoven (2002). An example of this type (morpheme
gloss mine) is shown in (28).
(28) Nuka-p puisi ame-er-paa [Kalaallisut from Van Geenhoven 2002, ex. (1)]
Nuka-rel. seal(abs) skin-remove-trans.3s/s
Nuka removed the skin from the seal
21There
must be something more to this; otherwise it would predict that the

subject becomes a N.
20
As Van Geenhoven (2002) convincingly shows, these verbs do not involve possessor
raising but are instead are verbs which seemingly have three arguments, an agent,
theme (incorporated) and a goal. In other words, the more literal translation of (28) is
the one shown, and not Nuka removed the seals skin, as would be the case under a
possessor raising analysis.22
Under the account of NI presented here, verbs of removing and providing would
seem to be a problem for a light verb analysis; therefore there must be an
explanation for why these verbs seem to be different. The answer to this lies in the
fact that these verbs are morphologically complex. Transitivization can create
causative forms from intransitive verbs in Inuktitut. While a full analysis of polyadic
NI verbs is beyond the scope of this paper, it can easily be seen that all verbs of
removing and providing are based on simpler forms, such as we have seen above.
Consider the data from the Inuvialuktun (Siglit) dialect.
(29) a. niaqu-iq-tuq
head-lack-intr.part.3s
It has no more head.
b. niaqu-i-gaa
head-lack-tr.part.3s/3s
He took its head off
(Inuvialuktun - Lowe 1984, 109)
(Inuvialuktun - Lowe 1984, 109)
The verb in (29b) is the same verb as in (29a). The change in meaning comes about
as a result of the transitive morphology following it. In some cases, the polyadic NI
verbs have applicative morphology as well, as the verb -liuut- meaning build for cited
in Van Geenhoven 2002, 761. This verb appears to be composed of -liuq- make plus
the applicative morpheme -ut(i)- (thus -liu-ut). What is crucial here is that the
derivational morphology appears to the right of the NI verbs, thus supporting the
claim made here that the NI verb on its own does not have any more content than a
light verb, i.e. is neither semantically nor syntactically complex.
We have achieved the goal as originally stated. We have seen that all noun
incorporating verbs are maximally simple semantically. In fact, all of them are
variants of have or be. Importantly, none of them involve manner of action e.g.
paint, brush, etc., change of state, e.g. open, break, etc. All such verbs are
independent non-incorporating verbs in Inuktitut.
While many of the NI verbs have the meaning of canonical light verbs, others
22It is not clear the degree to which the possessor raising analysis for these
constructions in Inuktitut is a straw man. Van Geenhoven (2002, 760) attributes it
to Bittner (1994). On p. 769 Van Geenhoven states that her discussion will be on the
construction would be analysed in a Bittner and Hale (1996) framework. In fact
Bittner (1994, 67) describes one of these constructions as containing a triadic
suffix, in keeping with Van Geenhovens subsequent analysis; however on p. 71
Bittner analyses a similar construction as pseudo-triadic or possessor raising.
21

can readily be shown to be related via simple elaboration of the semantics through
the addition of basic semantic operators. We have had to make use of three
operators: negation, quantification and identity. While identity and negation are quite
stable, always within the internal brackets, quantity plays a pivotal role through its
varying positions. This seems to be an inherent property of quantifiers. In the next
section, we will examine the syntactic structure of the construction containing an NI
verb and an incorporated noun.
3. THE SYNTAX OF LIGHT VERBS
As mentioned above, the incorporating verbs under discussion (which form the
majority of incorporating verbs in Inuktitut), have properties very similar to the
modal part of the bipartite lexical structure proposed in Koenig and Davis (2001). In
this section, I will propose an analysis which will explain why these verbs must have
a nominal attached to them. In this, I am addressing the important question raised
by Talmy (1985), which is how and why are verbs different in different languages (see
also Harley 2001 and Dchaine 2002). While adopting the modality division of Koenig
and Davis, I will instantiate it through the syntactic structures of Harley (2001).
Harley provides a typology of (English) verbs with four basic structures, below. The
first in (30) is the structure underlying unergative verbs, where X indicates the root
element which incorporates into little v, thus providing the verbs with a name.23
(30)
vP
2
v
X
The next type in (31) is the structure underlying do verbs like push, hit, etc.
(31)
vP
2
v
XP
3
X
YP
The third type is that shown in (32) which underlies cause type verbs, like transitive
clean.
23Harley has (agent) in the specifier position above these verbs, which I am
leaving out here. In Harleys framework, agency is indicated by putting an element in
spec vP.
22
(32)
vP
2
v
XP
3
YP
X
Finally the fourth type in (33) underlies give type verbs, like saddle, shelve, etc.
(33)
vP
2
v
XP
3
YP
X
3
X
ZP
For each verb type in English, the X incorporates into v, such that the root ends up
naming the verb as a whole, even though the structure from which it moves is
different. As a result, each little v in English will have a different derivational history.
According to Harleys analysis, there are no different light verbs, just different
structures as complement to little v.
Clearly as far as the analysis we have seen of incorporating verbs in
Inuktitut is concerned, NI verbs fall solely under type a) except that the X is always a
nominal and, most importantly, there is no incorporation into little v. Indeed what
characterizes Inuktitut NI verbs from this perspective is their very lack of
incorporation. They are pure little v, with the possible additions of the operators we
have seen above. That there is no incorporation and that the complement of little v is
a nominal may be related. Since no incorporation takes place, the complement of
little v must be an element which is independently licensed. Within Distributed
Morphology, a root must be licensed by a functional morpheme (Marantz 1997). It
seems that a little v can only license a complement either a) if the root of the
complement incorporates into little v or b) if the root is already incorporated into
another little v, as in the case of verbs embedded under morphological causatives. In
other words, little v cannot not license bare roots in complement position. A bare N
must be licensed then either by incorporating into little v, as in the case of
unergatives, or by being licensed by something else. The only other possible licenser
for a bare N in English is a D. Recall that above it was claimed that Inuktitut NPs
are inherently licensed, i.e. are effectively DPs, although they are not phrasal
syntactically. Thus a nominal can appear as the complement of little v, and need not,
indeed cannot (if it is already licensed), incorporate.24 Inuktitut NI constructions are
24As to why there must be morphological attachment, this question is part of
the larger question in Inuktitut as to why the majority of light lexical elements are
morphologically bound, e.g. adverbs, negation, aspect, tense, modals, etc. In other
23

base-generated with a bare nominal root (Ghomeshi and Massam 1994; Van
Geenhoven 1998a, 1998b; Massam 2001), through the operation Merge (Chomsky
2001). In fact, Hale (2001) claims that all noun incorporation universally involves
Merge, and certainly Inuktitut supports this claim. The result of merging a nominal
root and light v is a complete predicate, but one whose components remain distinct at
spell-out, as in (34), which is the structure underlying all NI verbs in Inuktitut.25
(34).
vP
3
v
N(=DP)
From this we see that there is no lexical verb root in this structure, and that while the
little v is head of the vP or verb-phrase, the nominal complement has a slightly
different relation to the vP than it would have if there were a full lexical verb in the
structure. In fact, from a certain perspective, the nominal IS the lexical content of
the verb. Harley (2001) proposes essentially the same structure in for verbs of
creation/consumption, as shown in (35).
(35)
vP
WRITE
3
v
DP
!
the book
John is writing the book
The structure in (35) represents the subject, which is generated as the specifier of
little v. The Incremental Theme verb is identical to unergative verbs (see 30 above),
except that there is no incorporation. In (35) there is also an incorporated manner
WRITE which originates from a position on the right. Exactly as in the account
argued here, the theme (the book) is a complement to little v. As Harley (2001, 14)
states, incremental them objects are direct objects of a light verb of creation (or
negative creation).
words, once we have designated NI verbs as light, the fact that must be
morphologically attached is no longer a matter of stipulation but part of a more
general issue in the language.
25Since the root is to the left of the light verb, there must also be some other
process which produces the correct morpheme order. Note that this is not noun
incorporation because all roots must be the leftmost element of the word. Thus even
a canonical verb will have a causative to its right. I leave aside this issue as it is
orthogonal to the paper.
24

We now have arrived at the answer for why the incorporating verbs in
Inuktitut are both a finite class, and semantically so general in meaning. Since the
verbal component of the construction is only a light v, there is a restriction on the set
of semantic contrasts possible. A verb which involves change of state, manner, etc.
would necessarily require that there be incorporation into little v. NI verbs in
Inuktitut have no incorporation whatsoever.
A by-product of this analysis is the fact that it predicts that an external
argument can never appear as an incorporated element, since it will always be
merged after the vP is created (see also Van Geenhoven 2002). Like the analysis in
Van Geenhoven (1998a), the analysis here argues for a base-generated construction,
involving no movement of a nominal from an object position; however unlike Van
Geenhovens analysis, and unlike all previous analyses of Inuktitut noun
incorporation (with the exception of Mithun 1997; 1999 - see below for discussion), the
present account explains the most salient features of the construction: a) only some
verbs incorporate and b) incorporation is obligatory. A) follows from the fact that
the semantics of pure light verbs is intrinsically restricted and b) follows from the fact
that in Inuktitut a bare N does not need an independent licenser, and that a light verb
belongs to a class of elements which require a root.
The light verb account of noun incorporation in Inuktitut predicts that in a
language with obligatory NI, noun incorporation will be a restricted phenomenon, light
verbs being only a subset of the entire class of verbs. Previous accounts, not
addressing the fact that only some verbs incorporate, have made incorporation a
special morphological feature on the lexical item, e.g. Bok Bennema and Groos (1988).
Van Geenhoven 2002, 808 states that For purely morphological reasons, an affixal
verb requires a nominal stem. Such accounts provide no explanation for a) why NI
verbs are not the general case , i.e. the majority of verbs, and b) why it is that a wide
range of possible meanings, e.g. tickle, shatter etc. are impossible and, as I claim
here implausible, incorporating verbs in Inuktitut.
Moreover the account of noun incorporating verbs as light verbs provides
empirical evidence for the existence of little v as an independent unit in syntax and
semantics, as well as support for Harleys (2001) approach to the lexical semantics of
verbs class. In particular, it confirms the claim that verbs of creation and
consumption are light verbs. In fact the class of NI verbs in Inuktitut is a robust set
of data which sheds light on the nature of light verbs, since they can be readily
identified by their having to appear with a merged nominal on the left.
4. IMPLICATIONS BEYOND INUKTITUT
The phenomenon described here, whereby restrictions on noun incorporation are
shown to be the direct result of the fact that the verbs involved are all light verbs is
found in other languages as well. Gerdts and Hukari (2002) describe what they term
denominal verbs in Halkomelem Salish, where the verbs are a) obligatorily attached
to a nominal stem and b) are a restricted set. A Halkomelem example is shown in
(36).
25

(36)
kwT John
ni/ txw-ka:
aux buy-car
det
John
John bought a car
The meanings buy is found among the Inuktitut light incorporating verbs, usually
combined with the meaning get or acquire. Not surprisingly,the other incorporating
verbs mean have/get/make/do, ingest/partake. They all create intransitive verb
forms, just as the examples in Inuktitut examined above do. These verbs are clearly
broad in meaning, as expected under the light verb analysis. More importantly, none
of their meanings contains root elements of manner or change. Nevertheless
Halkomelem and Inuktitut NI is not completely identical. Halkomelem does not allow
stranded modifiers while Inuktitut does (see Sadock 1980; 1991 and Van Geenhoven
1998a for an analysis of this), and Halkomelem allows object doubling, where the
same nominal appears both inside and outside the verbal construction. It is likely
that these differences derives from independent differences in the languages.
Crosslinguistically, noun incorporation involving denominal verbs is likely to
fall under the light verb analysis. Gerdts and Hukari (2002) cite the language Seri as
containing only a single denominal verb. It means have/put on/have as. These three
meanings are also found in Inuktitut NI verbs. The coincidence is striking, and
supports the light verb account.
5. ORIGINS OF NOUN INCORPORATING VERBAL AFFIXES
The analysis presented here claims that noun incorporating verbs of the type found in
Inuktitut and in other languages (sometimes called denominal verbs or verbal affixes)
are intrinsically light verbs only and that their restricted set of meanings and the fact
that they obligatorily incorporate are explained from this fact. However other
linguists have claimed that verbs of this sort result from grammaticalization
processes (Mithun 1997; 1999 and Gerdts and Hukari 2002).
Under a grammaticalization account, the NI verbs would be semantically
simple because of semantic bleaching. Similarly, the requirement that they must
have a noun attached would be explained by the fact that grammaticalization has
turned them into suffixes. The claim therefore is that they originated as full verbs and
due to frequency of use underwent grammaticalization into affixes.26
Mithun (1997; 1999) discusses prefix verbs in Salish, such as those we saw
above in Halkomelem. She observes two properties a) that they are similar in
meaning to verbal affixes in other languages and b) that they are very general in
meaning. She states p. 367 that most of these verbs simply indicate the presence or
absence of entities, their coming into being, their appearance on the scene. She also
26Malouf
(1999) also treats NI verbs in West Greenlandic as suffixes within an

HPSG account; however he stipulates that the NI verbs cannot be independent by
labelling them as verbal-bases, thus ignoring semantic generalizations. In addition,
his claim that the verbal-base can have the full valency of regular verbs, predicts a
huge number of impossible occurrences (see for example footnote 8.).
26

notes that many of the English translations show specific meanings, but that these
are a result of the nominals to which they are attached, not from the verb itself. This
parallels exactly what we saw in section 2.4.3. with the verb -tuq- consume in
Inuktitut, which can variously mean eat or drink depending on the nominal to the
left. Mithun also points out that the Salish class of verbs is similar in meaning to the
class of incorporating verbs in Yupik Eskimo.27 Significantly, Mithun (1997, 368)
mentions that the latter show a high proportion of markers indicating presence or
absence. This is the antonym-like property discussed above in section 2.3.1. In fact
Mithun seems to have noticed a great number of the generalizations which have been
the basis of discussion in the above sections, and they are well-founded. However her
conclusions are very different from the proposal in this paper. She posits that Yupik
(and Salish) verbal affixes (NI verbs) are the diachronic descendants of independent
lexical verbs which have been grammaticalized over time.
Mithun sees problems with this claim however. She ponders the question as to
why, if grammaticalization has taken place, these elements have retained more rootlike (lexical) properties than other affixes in the language. The root-like properties are
a) that they are a numerous, even though a closed class, and b) that they are have
concrete meanings relative to other affixes.28 Her answer to this problem is the
proposal that the historical path of these incorporating verbs differs from other
grammaticalization processes in that each verb did not undergo grammaticalization
independently, but instead an entire class of compounded forms underwent
grammaticalization, presumably around the same time. Mithun (1999) proposes a
grammaticalization history for Inuktitut noun incorporation based on earlier
compounding, even though she acknowledges that there is no compounding in the
language. Indeed there is neither synchronic nor diachronic evidence for compounding
in Inuktitut. Besides the lack of evidence for compounding, another problem with the
compounding source for NI verbs is that it is based on the concept of recurring
activities as the trigger for the compounding/grammaticalization. The fact that
virtually any pragmatically possible noun can appear with an NI verb does not seem
to be in keeping with a process determined by frequency of use.
Michael Fortescue , who is the foremost expert on the historical linguistics of
Eskimo-Aleut, explicitly claims that affixes in Eskimo languages do not draw upon
lexical stems, but from each other, through morphophonological change and recurrent
lexicalization across affixes.29 Fortescue (1992, 8) states that none of the affixes can
be related to lexical bases, despite the lexical weight of many of them.
The proposal in this paper does not assume that NI verbs were ever anything
other than what they are now, therefore no grammaticalization accounts would be
27What
does seem to be different between Inuktitut and Salish languages is

that in the latter some verb-noun combinations are lexicalized.
28We might also wonder why the grammaticalization is so uniform, each verb
have reached the same degree of semantic bleaching.
29Sometimes it is difficult to know whether or not lexicalization has taken
place, as in the case of the NI verb -liuut- in Kalaallisut discussed in section 2.4.4.
27

posited.30 The fact that compounds do not exist, nor presumably ever existed, in
Inuktitut is of no import to the light verb analysis. Likewise the fact that the class of
nominals which are merged with the light verbs are a large, open class, is exactly as
expected if the nominals are merged during the derivation to form the lexical root of
the vP.
5.1 The Stability of Light Verbs
As argued in section 5.0 there is no evidence to assume that the incorporating verbs
in Inuktitut are the result of grammaticalization process. If grammaticalization had
taken place, even if it were over an entire class as proposed by Mithun (1997; 1999),
we might expect to see some comparative differences across dialects. Instead we
observe that this class of verbs is quite uniform throughout the class, e.g. no verb
carries manner as part of its lexical semantic inventory. Equally we see exactly the
same range of verbs across the many dialects of Inuktitut from Alaska to Greenland.
The class of NI verbs in Inuktitut seems to be a reasonably stable across dialects of
Inuktitut, and potential membership in the class can be predicted based on lexical
meaning of the item. We do not find any semantically bleached versions of chew or
tickle in the language. Even common traditional Inuit activities, such as skin, cook,
sew or clean are not found as affixes.
Butt and Lahiri (2002) discuss light verbs in South Asian languages. In these
languages, the class of light verbs originally evolved from appearing after a verbal
noun to appearing after a bare verb stem, as in the Urdu example in (37).
(37)
nadya
a
ga-yi
Nadya
come go-Perf.F.Sg
Nadya has arrived
In (37) the light verb ga meaning go is added for semantic modification of the event of
Nadyas coming, not to indicate that Nadya went anywhere.
The phenomenon in South Asian languages is different from NI in Inuktitut, in
that the complement of the light verb is a verbal element and, also that each light
verb also has a full verb lexical counterpart which is identical in form. Butt and Lahiri
consider light verbs to be V-V, Adj-V or N-V constructions. Importantly they examine
the history the V-V construction in South Asian languages, arguing that it has been
present in the language since the earliest written records (estimated at possibly 1200
BCE). They claim that this fact is due to the light verb construction being a basic
part of universal grammar, rather than some construction which has evolved through
grammaticalization. Their central is that as far as historical change is concerned,
light verbs are a dead end, and that this fact derives from the central role they play
30If it turns out that grammaticalization has in fact occurred in Salish
languages, it may be that the verbs did not become prefixes until a certain level of
bleaching has occurred.
28

in universal grammar31. This is exactly what is being claimed here. NI verbs have the
properties which they do because they form a special subclass of verbs in human
language, a subclass with distinct syntax and distinct semantics. Their distinct
syntax is that they are associated with little v and therefore, if no other material is
incorporated, have no root properties. Their distinct semantics is that they have no
manner or change of state, thus presenting as a semantically bleached set of verbs
as we have seen above in section 2. There is no need to derive these verbs from other
verbs. They themselves form the initial building blocks of all verbs (see Harley 2001).
6. CONCLUSION
We have seen that a close look at the subclass of verbs involved in noun
incorporation in Inuktitut is very revealing. Previous analyses of this phenomenon in
this language have ignored this issue, with the exception of Mithun (1997; 1999). It
turns out that the subclass in question can readily be shown to consist of meanings
which are associated with light verbs or little v. In Inuktitut, many of these verbs
sometimes also have operators of negation, quantity, etc. As a result, they seem
semantically related to one another, almost as if in sets. This explains the limited
semantic range which we find in these verbs. The fact that these verbs are in little v
also provides an explanation for the fact that a nominal must be attached with the
little v, thus providing a root element. As a result, it is clear that noun incorporation
in Inuktitut does not involve movement (Van Geenhoven 1998a, b, 2002; Massam
2001) but simple Merge (Hale 2001).
As a result of this analysis we can dispense with all previous accounts which
attribute an arbitrary morphological feature to merged light verbs. These verbs
obligatorily suffix to the nouns because they are not root elements themselves.
However, they are not affixes as a result of grammaticalization, as claimed in Mithun
(1997; 1999). There is no plausible account along these lines in Inuktitut. Instead, the
light verb account explains these properties in a straightforward fashion. Light verbs
are a universally available subclass of verbs (Harley 2001; Butt and Lahiri 2002)
That other languages with a similar phenomenon show members from the same
subset supports this account.
This analysis supports a bipartite structure of verb syntax and semantics
along the lines of Harley (2001) (see Koenig and Davis 2001 for a lexical semantic
proposal). Verbs within and across languages vary depending on what sort of
complements the little v takes and which elements incorporate (see also Dchaine
2002). In the case of Inuktitut, incorporation does not take place with a certain
subclass of verbs, therefore these verbs show special properties within the language.
Thus the light verb account affords us deeper understanding into the nature of verb
types, and the ways in which languages are able to instantiate these differences.
31They
note that universally verbs such as take put make and do are
semantically underspecified.
29
REFERENCES
Baker, Mark. 1988. Incorporation: A Theory of Grammatical Function Changing,
University of Chicago Press, Chicago.
Baker, Mark. 1996. The Polysynthesis Parameter, Oxford University Press, New York.
Bittner, Maria. 1994. Case, Scope, and Binding, Kluwer Academic Publishers, Boston.
Bok-Bennema, Reineke and Anneke Groos. 1988. Adjacency and Incorporation, in .
Martin Everaert, et al. (eds.) Morphology and Modularity, Foris, Dordrecht, pp. , 3356.
Butt, Miriam and Aditi Lahiri. 2002. Historical Stability vs. Historical Change
manuscript, University of Konstanz.
Chomsky, Noam. 2001. Derivation by Phase, in M. Kenstowicz (ed.) Ken Hale: A Life
in Language, MIT Press, Cambridge, MA, pp. 1-52.
Dchaine, Rose-Marie. 2002. On the Significance of (Non) Augmented Roots. Paper
presented at the annual meeting of the Canadian Linguistic Association, Toronto.
Fortescue, Michael. 1983. A Comparative Manual of Affixes for the Inuit Dialects of
Greenland, Canada, and Alaska, Meddelelser om GrPnland, Copenhagen.
Fortescue, Michael. 1992. The Development of Morphophonemic Complexity in
Eskimo, Acta Linguistica Hafniensa 25, 5-27.
Gerdts, Donna. 1998. Incorporation, in Andrew. Spencer and Arnold Zwicky (eds.),
Handbook of Morphology, Basil Blackwell, London, pp. 84-100.
Gerdts, Donna, and Thomas Hukari. 2002. Doubling and Denominal Verbs: Evidence
from Halkomelem Salish, Paper presented at the annual meeting of the Canadian
Linguistics Association.
Ghomeshi, Jila and Diane Massam. 1994. Lexical/Syntactic Relations without
Projection, Linguistic Analysis 24, 175-217.
Grimshaw, Jane and Armin Mester. 1988. Light Verbs and Theta Marking,
Linguistic Inquiry, 19:2, 205-232.
Hale, Kenneth. 2001. Navajo Verb Structure and Noun Incorporation, Ms. MIT.
Halle, Morris and Alec Marantz. 1993. 'Distributed Morphology and the Pieces of
Inflection' , in Kenneth Hale and S. Jay Keyser (eds.), The View from Building 20
MIT Press, Cambridge, pp. 111-176.
Harley, Heidi. 2001. Measuring-out and the Ontology of Verb Roots in English. Paper
presented at the Ben-University Workshop on Aspect, June 18-20, 2001.
Jackendoff, Ray. 1990. Semantic Structures, MIT Press, Cambridge.
Jespersen, Otto. 1965. A Modern English Grammar on Historical Principles, Part VI,
George Allen and Unwin Ltd., London.
Johns, Alana. 1987. Transitivity and Grammatical Relations in Inuktitut,
unpublished Ph.D. dissertation, University of Ottawa, Ottawa.
Johns, Alana. 2000 Restricting Noun Incorporation. Paper presented at the annual
meeting of the LSA, Chicago, Jan. 2000.
Johns, Alana. 1999a. The Lexical Basis of Noun Incorporation in Inuktitut. Paper
presented at the Workshop on Structure and Constituency of the Languages of the
Americas. UBC.March 1999.
30

Johns, Alana. 1999b. On the Lexical Semantics of Affixal want in Inuktitut.
International Journal of American Linguistics 65:2, 176-200.
Johns, Alana and Diane Massam. 1998. Questions Raised by Noun
Incorporation.Paper presented at the Third Workshop on Multiple Frameworks in
Linguistics, University of Toronto.
Kearns, Kate. 1988. Light Verbs in English, Ms, University of Cantebury, New
Zealand.
Koenig, Jean-Pierre and Anthony Davis. 2001. Sublexical Modality and the Structure
of Lexical Semantics, Linguistics and Philosophy, 24:1, 71-124.
Lowe, Ronald. 2001. Siglit Inuvialuit Uqautchiita Nutaat Kipuktirutait Aglipkaqtat
(Siglit Inuvialuit Eskimo Dictionary), ditions Nota Bene,Qubec.
Malouf, Robert. 1999. West Greenlandic Noun Incorporation in a Monohierarchical
Theory of Grammar, in Gert Webelhuth, Andreas Kathol, and Jean-Pierre Koenig
(eds.), Lexical and Constructional Aspects of Linguistic Explanation. CSLI
Publications, Stanford, pp. 47-62.
Marantz, Alec. 1997. No Escape from Syntax: Don't Try Morphological Analysis in
the Privacy of Your Own Lexicon, in Alexis Dimitriadis et. al. (eds.) University of
Pennsylvania Working Papers in Linguistics, vol. 4.2, pp. 201-225
Massam, Diane. 2001. Pseudo Noun Incorporation In Niuean, Natural Language
and Linguistic Theory 19:1, 153-197.
Mithun, Marianne. 1984. The Evolution of Noun Incorporation. Language 60, 847893.
Mithun, Marianne. 1997. Lexical Affixes and Morphological Typology, in Joan Bybee
et al. (eds.) Essays on Language Function and Language Type, John Benjamins,
Philadelphia, pp. 357-371.
Mithun, Marianne. 1999. The Languages of North America, Cambridge Language
Surveys, Cambridge University Press, New York.
Mithun, Marianne. and Greville Corbett. 1999. The Effect of Noun Incorporation on
Argument Structure, in Lunella Mereu (ed.) Boundaries of Morphology and Syntax,
John Benjamins, Amsterdam, pp. 49-72.
Parkinson, David. 1999. The Interaction of Syntax and Morphology in the Acquisition
of Noun Incorporation in Inuktitut, unpublished Ph.D. dissertation, Cornell
University, Ithaca.
Pesetsky, David. 1995. Zero Syntax. MIT Press, Cambridge.
Rappaport Hovav, Malka, and Beth Levin. 1998. Building Verb Meanings, in Miriam
Butt and Wilhelm Geuder (eds.) The Projection of Arguments: Lexical and
Compositional Factors, pp. 97-134. CSLI, Stanford
Ritter, Elizabeth. and Sara Thomas Rosen. 1997. The Function of have, Lingua 101,
295-301.
Rosen, Sara Thomas. 1989. Two Types of Noun Incorporation: A Lexical Analysis,
Language 65, 294-317.
Sadock, Jerrold. 1980. Noun Incorporation in Greenlandic: A Case of Syntactic Word
Formation, Language 56, 300-319.
Sadock, Jerrold. 1985. Autolexical Syntax: A Theory of Noun Incorporation and
Similar Phenomena, Natural Language and Linguistic Theory 3, 379-440.
31

Sadock, Jerrold. 1991. Autolexical Syntax, University of Chicago Press, Chicago.
Sadock, Jerrold. 2002. A Survey of Denominal Verbs in Eskimo-Aleut. Paper
presented at the SSILA Meeting, San Francisco.
Sapir, Edward. 1911. The Problem of Noun Incorporation in American languages,
American Anthropologist 13, 250-282.
Talmy, Leonard (1985) Lexicalization Patterns: Semantic Structure in Lexical
Forms, in Timothy Shopen (ed.), Language Typology and Syntactic Description III:
Grammatical Categories and the Lexicon, pp. 57-149.
Tenny, Carol. 1987. Grammaticalizing Aspect and Affectedness. Unpublished Ph.D.
thesis, MIT, Department of Linguistics and Philosophy.
Van Geenhoven, Veerle. 1998a. Semantic Incorporation and Indefinite Descriptions:
Semantic and Syntactic Aspects of Noun Incorporation in West Greenlandic,
Dissertations in Linguistics, CSLI Publications, Stanford.
Van Geenhoven, Veerle. 1998b. On the Argument Structure of some Noun
Incorporating Verbs in West Greenlandic, in Miriam Butt and Wilhelm Geuder
(eds.) The Projection of Arguments: Lexical and Compositional Factors, CSLI,
Stanford, pp. 225-263.
Van Geenhoven, Veerle. 2002. Raised Possessors and Noun Incorporation in West
Greenlandic, Natural Language and Linguistic Theory 20, 759-821.
32

Restricting Noun Incorporation

Transféré par

Informations du document

Copyright

Formats disponibles

Partager ce document

Partager ou intégrer le document

Options de partage

Avez-vous trouvé ce document utile ?

Ce contenu est-il inapproprié ?

Droits d'auteur :

Formats disponibles

Restricting Noun Incorporation

Transféré par

Droits d'auteur :

Formats disponibles

RESTRICTING NOUN INCORPORATION

Alana Johns, University of Toronto (ajohns@chass.utoronto.ca)

(Mohawk, Baker 1996 p. 279)